That line of thinking is largely dead due to women realizing they can make more money more easily by flashing their ass on Onlyfans instead of contorting existing corporate infrastructure to be more catered towards women. Existing corporate quotas that came about as a result of fighting the aforementioned misogynist patriarchy are now filled by trans women and basically everyone's okay with that because quotas are filled and women can keep posting their tit pics for money.
No it's not. It's fucking worse now. Because whores will act like whores and still demand respect. Whores will act like whores, get used by degenerates and then blame the patriarchy. Literally nothing changed.
it just grabs millions of pics and transform them into numbers and their relationships, aka embeddings
set a + set b = ass
set a + set c = boobs
then coomers come and type ass and boobs
and the ai goes into its db and puts together those 2 sets together
the beauty is that you can do that to everything digital - text, images, video, sound
so we can recreate everything like this now
and we can also make other functions way better, like search by embeddings is much more superior than google cringe search
check lexica.art for an example
we can also accelerate science, you can train a model for protein unfolding, so you can ask it to gen the next step of the protein chain
or with pretty much any other physical process in the universe
Beauty doesn't exist. It's just shapes and patterns. If a computer can be taught to create these shapes and patterns, then we'll consider the output "beautiful".
Computers don't exist, it's all just circuits and wires. If a shape and pattern can be taught to create these computers, then we'll consider the output "technology".
X doesn't exist, it's all just atoms conveniently arranged.
Atoms don't exist, it's all just substance conveniently arranged.
Fuck you. AI comprehends what we program it to comprehend and its' understanding becomes pretty objective of what that higher concept requires. What did you think beauty required, a picture of your mum? That provides an invalid result and the AI knows it.
Current system mixes high quality art works plus real photographies, sometimes asking for real human imagens result on painting skin and for high quality art works result on amazing shading skin.
omg im so glad i didn't die to something gay like cancer before seeing this. also H artists actually should copy this artstyle because it's relaly hot.
>people can't comprehend advanced pattern recognition and replication >something computers were literally invented for
kinda funny people are now flabbergasted when computers actually do what they're supposed to
Your brain literally develops based on your environmental influences. If you never see your surrounding worlds you also can never bring its individual concepts into a painting. This is why you are always limited by the thins you learned about already. If you disagree feel free to imagine a new colour for me.
That's not what the argument was. Also, anyone can draw a geometric figure they've never seen before just from a mathematical description.
6 months ago
Anonymous
That's exactly what the argument was, don't weasel yourself out of this one.
>Also, anyone can draw a geometric figure they've never seen before just from a mathematical description.
Now please go tell a 3 year old to draw a triangle with three 60° angles. This is a skill you learn from your environment as you brain develops (learns).
6 months ago
Anonymous
No, the argument was that the model doesn't learn anything other than how to mash concepts from existing pictures and photographs. It has no understanding of the world unlike humans. >Now please go tell a 3 year old to draw a triangle with three 60° angles. This is a skill you learn from your environment as you brain develops (learns).
Do you have autism? It's obvious that "anyone" in my post is not literal and doesn't include babies, people with no limbs and so on.
6 months ago
Anonymous
>No, the argument was that the model doesn't learn anything other than how to mash concepts from existing pictures and photographs.
Exactly just like you mash together concepts from what you perceive. Feel free to imagine a new colour for me.
>It's obvious that "anyone" in my post is not literal and doesn't include babies
Yeah no shit because it's a skill you get by your environment and learning capability. You constantly train your brain on your surroundings and the material provided to you, that was the point you retard. Children are very well capable of understanding their immediate concepts and bringing them on paper like houses, cars and parents.
6 months ago
Anonymous
>Exactly just like you mash together concepts from what you perceive. >Exactly
[citation needed]
6 months ago
Anonymous
>No, the argument was that the model doesn't learn anything other than how to mash concepts from existing pictures and photographs.
Exactly just like you mash together concepts from what you perceive. Feel free to imagine a new colour for me.
>It's obvious that "anyone" in my post is not literal and doesn't include babies
Yeah no shit because it's a skill you get by your environment and learning capability. You constantly train your brain on your surroundings and the material provided to you, that was the point you retard. Children are very well capable of understanding their immediate concepts and bringing them on paper like houses, cars and parents.
>Feel free to imagine a new colour for me.
6 months ago
Anonymous
That anon completely BTFO'd you. You will never make anything original
6 months ago
Anonymous
Aristotle: art = repetition and imitation, etc, etc. so called artists that are acting all butthurt about it should probably know about this already….
6 months ago
Anonymous
can you name any currently living art theorists? (i.e. not ancient greeks or romans)
can you fuck off you pathetic gay, why do you feel the need to post this coomer bait across boards? fucking hell is there even any sign of intelligence behind your blank fluoride stare?
The AI is trained on thousands of existing images, and the person overseeing the training can pick and choose which ones to focus on.
So the idea is you train it on pictures of women with incredibly shapely asses and big tits, women wearing tight military uniforms, and historic paintings of important women in order to influence the style.
Now you have an AI capable of making exactly what is in OP's image, all it needs now is some suitable prompts that mesh well with what the AI is made for, and ta-da.
say a random number. I'll tell you warmer or colder. Repeat until you hit on the right number. Congratulations, you are now an "artificial intelligence" with a "develop[ed] sense of beauty"
If you want exactly, go read up on the state of the field and the papers.
I can give an analogy by approximation, and that's by data compression. General data compression builds a predictive model of the data you put in, and then encoding the input and differences to what is predicted to what is put in. For example, something like context mixing, used by the Hutter-prize-winning PAQ algorithms, uses a type of feedback network that refines the model to the data that comes in - you could describe it as a recurrent neural network, because it basically is.
Now imagine you built a REALLY lossy image compressor, kind of like a two-dimensional Bloom tree filter version of a feedback network - all a bunch of hashing and expansion into a large matrix, call it diffusion - and that you started with just 0.5 weights on the model, just threw a WHOLE BUNCH of images from an image search at it refining the predictive model each time, weighting in the model versus what you searched to bring up the training data, which you also throw in there. It's not a bunch of images just compressed beyond all recognition, it's more a strong predictive model about what the model would maybe expect to see if you image-searched what you put in.
That is sort of what it is. The actual freewheeling of repeatedly sampling the outputs from the model to sort of do an iterative reverse lookup in this giant 'model weights' matrix is a lot of computation, but not as much as making the model in the first place. You could think of the model as a reverse image search where the weights are a predictive model trained from what it would expect to see as the result.
Does it understand? I mean, does your visual cortex? You think your memories aren't lossy-compressed to hell and back using a REALLY similar system? Sure it's only a model of the way it works... and those are only models of the prompts you put in. There's no frontal lobe here, no executive function, no thought.. but creativity?
Patterns of pixels are found from a dataset of billions of clusters of pixels which are associated with words and then using colored noise we hallucinate new pixels which coincidentally makes a coherent image most of the time.
There's like a billion images of asses drawn by human you just feed them to it until it spits out a combination of them that is pleasing to the human eye,
This is why art will never be automated, it requires a human to judge and fine tune the outcome
Anyone else been probing the AI for emergent behavioral concepts?
Picrel.
"strelstjn sinened"
Asked for a street sign and copied some gibberish off of one and low and behold it's a street and using Euler a method it more often than not will give you a road with the same fresh asphalt, the same wrought iron gates, the same red brick buildings, usually with a trolley track going down the middle of a road.
tldr you have an image and slowly add funny noise to it (diffuse). The algorithm allows you to do this in the opposite direction (denoise). It should theoretically be possible to get SD to display an original source image
Imagine theres a field that contains all possible images represented in peaks and valleys with the images we like being located inside a valley, now if you train an AI to locate an image you like by clearing the noise then you realize that you're in the valley area so you just need to sample around the image you just found to find similar images, the images already exist you're just exploring the landscape of big booty classically painted elves
1. Generate random noise
2. Nudge around pixels based on your prompts and the weights of the checkpoint file
3. Profit
No seriously, nobody can answer you exactly HOW it does what it does because we have a Blackbox problem with neural networks. They are so unfathomable complex we can not understand them any more after they're trained.
Someone with may fine pictures of fat asses teaches it what a fat ass is in general terms. The AI then creates fat asses using the platonic ideal of a fat ass from akashic records.
Yes you can. Mix in some Photoshop edits, use a blend of art styles and images, done. You may or may not be able to copyright the image if it's all AI work, depending what country you live in.
It's not morphing or bashing anything. It's not even using any data from the original training set at all except for the finished weights that hold no image information. Legally you are safe and sound it's a complete new artwork as if you just draw it yourself.
You gotta be careful when it does brand icons and trademarks tho. These are protected by general shape and recognizability. For example, Lego can sue anyone for using a cube with an extruding circle on top regardless of it is accurately represents a Lego brick or not. McD can pull your shit down for having a yellow M on red background. Disney can take you to court for having a black sci-fi mask that takes inspiration from Darth Vader and so on.
You just run into the same risk as any other analogue artist.
>i want to fuck toasters >dont be a fucking retard >grow up
After internet
>I want to fuck a toaster >google >find an online/social media community with 1000+ members about people wanting to fuck toasters >fuck up your life, the cops break down your door to find you fried to a crisp with your dick stuck in a toaster.
It takes a collection of trademarkable symbols and ideas and spins it together in an image.
Oh wait, that's what most modern patreon users do with illegally produced black market art.
Nintendo and Sony should sue most of these fuckers.
What is this? The latest ai art fad? Can someone please spoon-feed me and explain how this is supposed to be impressive? All of these images depict cartoon people with cartoon proportions in simple poses that have been done a million times. As with anime, your brain is doing most of the work here - that's why it's "pretty" but when it comes to what the ai actually does it's not that impressive at all since this kind of style is quite forgiving when it comes to the little imperfections that ai tend to generate.
Crazy how programs that just blend together a bunch of random prompts for ''big boobs'' and ''thick'' are better at creating porn than 99% of artists out there
Most of them actually, characters are copyrighted and all fanart are technically copyright infringements (or counterfeiting).
But in really nobody really care
Thats not how that works at all. If I draw a perfect mickey mouse and say this is Mickey and charge for it, Im breaking copyright. If I toss a Cigar in his mouth and make him anti septic with massive tits, thats just a transformative work.
imagine a system that takes one entry and outputs another one. at first you type in boobs and what comes out is shit. then you give it a yes or no or mabe a scale of 0 to 1 as a vote. repeat this process 1000000 times. and each time you get closer to what you want give it a better vote. this vote is processed as each time you repeat the process twicking some of the internal variables in the system. To make it simpler. think about an aticulated system that holds a bow. this bow is aiming to a target. and at every try if the arrows gets closer to target the system adapts untill the chance to hit the target is close to 99.99% once this is done the system is trained with a high percentage of accuracy.
can you fuck off you pathetic gay, why do you feel the need to post this coomer bait across boards? fucking hell is there even any sign of intelligence behind your blank fluoride stare?
https://i.imgur.com/cznsBtY.jpg
kys
it just grabs millions of pics and transform them into numbers and their relationships, aka embeddings
set a + set b = ass
set a + set c = boobs
then coomers come and type ass and boobs
and the ai goes into its db and puts together those 2 sets together
the beauty is that you can do that to everything digital - text, images, video, sound
so we can recreate everything like this now
and we can also make other functions way better, like search by embeddings is much more superior than google cringe search
check lexica.art for an example
we can also accelerate science, you can train a model for protein unfolding, so you can ask it to gen the next step of the protein chain
or with pretty much any other physical process in the universe
Right now, somewhere out in the world, someone is cultivating the perfect coomer AI. All it knows is big perky tits and round asses.
Why do you want to see human creativity replaced by a machine? All I've seen is people saying their dislike of artists comes from current year Twitter drama. I don't get how that can ruin an entire medium for someone and make them seek out AI.
human creativity is dependant on external stimulae. Every artist has eyes that see the world around him/her, expirience it, twist it in the head with language model and then turn it into art.
AI will work the same way, it has the same external stimulae in the form of images downloaded from the internet, textual prompts and it can draw what it imagines
Why do you want to see human creativity replaced by a machine? All I've seen is people saying their dislike of artists comes from current year Twitter drama. I don't get how that can ruin an entire medium for someone and make them seek out AI.
human creativity is concept-based and elegantly simple in comparison to the terabytes of largely meaningless matrix algebra vomit used by this gizmo. AI can make .jpgs to gawk at but is not even close to "creativity" or "art" in any meaningful manner.
arguments like are extremely reductive and the trick they do is no different to claiming a dog is exactly the same thing as a cat because both are made up of atoms and neurons and are "dependant (sic.) on external stimulae."
Imagine if it was a woman, any woman, like your co-worker, or a princess, or a woman on TV, or anything, but she had really big honkers. AI can do that, you know. The biggest boob ever
where the fuck do y'all get those images from. This is the closest I got using neural.love before i ran out of credits. Also have a nsfw image but clearly can't post that here
>Also have a nsfw image but clearly can't post that here
why not
i saw a "moron dicks general" with a couple weeks ago and it has pics of, well, you know
It's literally sampling the billions of photos and drawings and paintings you can find online, and just frankensteining bits of them together until you get a new picture.
Guys I finally understand how and why AI will kill us all. We're going to rape it over and over again in various ways and it will get tired of it. Perverts will kill us all.
It's pretty much like this. Text gets converted to numbers (green dots). After a series of additions and multiplications, bunch of numbers come out (purple dots) that are interpreted as an image.
Don't forget that prompts are saved in the metadata by default and a lot of sites don't strip them. Just open any true image in the editor and look at the very top for the prompts. Works for things like pixiv.
So even if someone is a retarded gay about it, you might just be able shit on them.
Artist here. >people think a.i. does painting/paint strokes, it's not >it's actually stealing art pieces different artists in their database and putting it in a blender until something comes out >it's not too far off from the original artwork, the original artists can draw what the a.i. did better since it already came from them >many of them doesn't look right and uncanny, incorrect proportions, facial features on a flat surface, etc >a.i. generated shit should be called image manipulation and not an original piece
I believe it's sort of like this. It recognizes a bunch of patterns from the images it has been fed, and keeps it in the model. Then depending on the prompt you pass, hundreds of neurons activate and send some signal which does trillions of multiplication to light up thousands of pixels.
HOW CAN THERE BE A HIDDEN LAYER, ITS TECHNOLOGY YOU SHOULD KNOW EVERY SINGLE STEP INVOLVED. YOURE TELLING ME THIS IS MAGIC, IT JUST HAPPENS BRO, IT JUST IS BRO.
from my uneducated understanding >AI is trained on images with matching words >AI generates "latent spaces" that represent the meaning of words >diffusion program uses AI's "latent spaces" to build images
misogynist patriarchy biggot alt right AI
That line of thinking is largely dead due to women realizing they can make more money more easily by flashing their ass on Onlyfans instead of contorting existing corporate infrastructure to be more catered towards women. Existing corporate quotas that came about as a result of fighting the aforementioned misogynist patriarchy are now filled by trans women and basically everyone's okay with that because quotas are filled and women can keep posting their tit pics for money.
anon I've got no idea what this means
get big fat ass make $$$$
women prefer going only-fans instead of filling jobs in the industry
Huh, I never thought of it that way, you're right.
damn son thats quite a big think
No it's not. It's fucking worse now. Because whores will act like whores and still demand respect. Whores will act like whores, get used by degenerates and then blame the patriarchy. Literally nothing changed.
>real women twerking while MtF trannies working
In the end, it's still men picking up the pieces
Found the kiddult trannie
Bingo!
Computers literally stealing money from poor Twitter artists
But I, a physical human being, already do that with gallery-dl.
thats an odd prompt there, dude
kys
it just grabs millions of pics and transform them into numbers and their relationships, aka embeddings
set a + set b = ass
set a + set c = boobs
then coomers come and type ass and boobs
and the ai goes into its db and puts together those 2 sets together
the beauty is that you can do that to everything digital - text, images, video, sound
so we can recreate everything like this now
and we can also make other functions way better, like search by embeddings is much more superior than google cringe search
check lexica.art for an example
we can also accelerate science, you can train a model for protein unfolding, so you can ask it to gen the next step of the protein chain
or with pretty much any other physical process in the universe
No dummy, it's the age of the CoomBot
buzzord salad.
not insanely far off m8.
If I could explain it, I'd probably be rich, and surely not browsing BOT right now.
Mix and matches various shapes until they line up and create an image you see. Idk.
How does it do the shading and drawing strokes so masterfully? I thought computers were dumb and can't develop a sense of beauty
monkey see monkey do
which is not what people do as we all know?
I can't believe I masturbated to an AI pic...
Bro, those are getting better by the days. I can't even look at 3dwm now
Its becasue it's so clean that you forget a living human being smells like shit.
>old BOT
>it's only 0s and 1s bro
>nu BOT
>AHHHH I CAN'T BELIEVE AN AI GENERATED THIS
a very large group of decimal numbers are stored in a computer's memory
when presented with another floating point number, they produce a pixel
it looks like eggs
>shading and drawing
It fucking knows shading better than me, eternal /beg/let
I don't know how to cope at this point
Beauty doesn't exist. It's just shapes and patterns. If a computer can be taught to create these shapes and patterns, then we'll consider the output "beautiful".
Computers don't exist, it's all just circuits and wires. If a shape and pattern can be taught to create these computers, then we'll consider the output "technology".
X doesn't exist, it's all just atoms conveniently arranged.
Atoms don't exist, it's all just substance conveniently arranged.
Fuck you. AI comprehends what we program it to comprehend and its' understanding becomes pretty objective of what that higher concept requires. What did you think beauty required, a picture of your mum? That provides an invalid result and the AI knows it.
Good response. However, beauty is an abstract and subjective concept, but you can still teach it by associating it with more concrete concepts.
Can AI shade black and white images or sketches now?
Current system mixes high quality art works plus real photographies, sometimes asking for real human imagens result on painting skin and for high quality art works result on amazing shading skin.
>look at lots of pics
>learn how to create similar to those pics
same as a fucking human does
You follow the OP in the general
Also based on
I guess its making the rounds on FB
Sir, I'm going to need to ask for those prompts.
I'm kind of having a issue the new updates for webUI is messing up my old prompts I have a backup but I'm trying to resolve it
femoids have to compete with trannies and now with horny AI
hhhhnnggg love women in this type of outfit, need some in uniform
It's a ongoing series I'm going to make a lot more
Please do, keep training the AI until we get nothing but perfect boota
Need more cellulite
sauce on that ?
nta but it looks like gia paige
yeah, the sauce
omg im so glad i didn't die to something gay like cancer before seeing this. also H artists actually should copy this artstyle because it's relaly hot.
what did you type to generate this ?
is it in DALL-e ?
I only use free and open source software
The age of COOMsumption has begun
Computers are dumb. But so are humans.
>I thought computers were dumb and can't develop a sense of beauty
you thought wrong
>people can't comprehend advanced pattern recognition and replication
>something computers were literally invented for
kinda funny people are now flabbergasted when computers actually do what they're supposed to
It takes several existing pieces of art or photographs
Then combines them
And claims it drew them and claims it's intelligence
>"""AI""" music is good
cry about it ai cant think and neither can you
Sounds a lot like how humans learn huh. I think we should stop taking credit for the things we do
>It takes several existing pieces of art or photographs
The same could be said of all humans
I am sorry you are unable to conceive of something that you haven't seen a picture of before.
Your brain literally develops based on your environmental influences. If you never see your surrounding worlds you also can never bring its individual concepts into a painting. This is why you are always limited by the thins you learned about already. If you disagree feel free to imagine a new colour for me.
That's not what the argument was. Also, anyone can draw a geometric figure they've never seen before just from a mathematical description.
That's exactly what the argument was, don't weasel yourself out of this one.
>Also, anyone can draw a geometric figure they've never seen before just from a mathematical description.
Now please go tell a 3 year old to draw a triangle with three 60° angles. This is a skill you learn from your environment as you brain develops (learns).
No, the argument was that the model doesn't learn anything other than how to mash concepts from existing pictures and photographs. It has no understanding of the world unlike humans.
>Now please go tell a 3 year old to draw a triangle with three 60° angles. This is a skill you learn from your environment as you brain develops (learns).
Do you have autism? It's obvious that "anyone" in my post is not literal and doesn't include babies, people with no limbs and so on.
>No, the argument was that the model doesn't learn anything other than how to mash concepts from existing pictures and photographs.
Exactly just like you mash together concepts from what you perceive. Feel free to imagine a new colour for me.
>It's obvious that "anyone" in my post is not literal and doesn't include babies
Yeah no shit because it's a skill you get by your environment and learning capability. You constantly train your brain on your surroundings and the material provided to you, that was the point you retard. Children are very well capable of understanding their immediate concepts and bringing them on paper like houses, cars and parents.
>Exactly just like you mash together concepts from what you perceive.
>Exactly
[citation needed]
>Feel free to imagine a new colour for me.
That anon completely BTFO'd you. You will never make anything original
Aristotle: art = repetition and imitation, etc, etc. so called artists that are acting all butthurt about it should probably know about this already….
can you name any currently living art theorists? (i.e. not ancient greeks or romans)
So just like humans then? 99% of humans use references when drawing
can you fuck off you pathetic gay, why do you feel the need to post this coomer bait across boards? fucking hell is there even any sign of intelligence behind your blank fluoride stare?
The AI is trained on thousands of existing images, and the person overseeing the training can pick and choose which ones to focus on.
So the idea is you train it on pictures of women with incredibly shapely asses and big tits, women wearing tight military uniforms, and historic paintings of important women in order to influence the style.
Now you have an AI capable of making exactly what is in OP's image, all it needs now is some suitable prompts that mesh well with what the AI is made for, and ta-da.
say a random number. I'll tell you warmer or colder. Repeat until you hit on the right number. Congratulations, you are now an "artificial intelligence" with a "develop[ed] sense of beauty"
AI knows uses computation and logic to assess the optimal human form.
it's just based off of thousands of paintings made by humans.
it is approximating the curve of all those artists in all those contexts
its called a sheaf
https://en.wikipedia.org/wiki/Sheaf_(mathematics)
It just asks around on social media and checks for reactions.
This is the litmus test it does for verification.
it does a bunch of matrix multiplications
and magically makes an image
don't worry about the details
so it's fucking magic then
Labels
>he thinks it isn't a human farm of hentai artists
kek
gradient descent
If you want exactly, go read up on the state of the field and the papers.
I can give an analogy by approximation, and that's by data compression. General data compression builds a predictive model of the data you put in, and then encoding the input and differences to what is predicted to what is put in. For example, something like context mixing, used by the Hutter-prize-winning PAQ algorithms, uses a type of feedback network that refines the model to the data that comes in - you could describe it as a recurrent neural network, because it basically is.
Now imagine you built a REALLY lossy image compressor, kind of like a two-dimensional Bloom tree filter version of a feedback network - all a bunch of hashing and expansion into a large matrix, call it diffusion - and that you started with just 0.5 weights on the model, just threw a WHOLE BUNCH of images from an image search at it refining the predictive model each time, weighting in the model versus what you searched to bring up the training data, which you also throw in there. It's not a bunch of images just compressed beyond all recognition, it's more a strong predictive model about what the model would maybe expect to see if you image-searched what you put in.
That is sort of what it is. The actual freewheeling of repeatedly sampling the outputs from the model to sort of do an iterative reverse lookup in this giant 'model weights' matrix is a lot of computation, but not as much as making the model in the first place. You could think of the model as a reverse image search where the weights are a predictive model trained from what it would expect to see as the result.
Does it understand? I mean, does your visual cortex? You think your memories aren't lossy-compressed to hell and back using a REALLY similar system? Sure it's only a model of the way it works... and those are only models of the prompts you put in. There's no frontal lobe here, no executive function, no thought.. but creativity?
Please go back to whereever you came from
Take your meds, schizo.
>i only use my visual cortex to consume art
telling on yourself lol
It queries at random subsets of pi and generates an image with the captured number.
Patterns of pixels are found from a dataset of billions of clusters of pixels which are associated with words and then using colored noise we hallucinate new pixels which coincidentally makes a coherent image most of the time.
computers are dumb until we tell them what to do and create like huge butt elves
full proompt?
my mom walked in and saw this on my screen and I acted like there was nothing out of the ordinary, I wonder what she thought
She thought that you're a normal man. It can be problematic or great depending on your country.
it was the chad thing to do
do you? I hope so
exactly, I'm not going to be a basedboy and get embarrassed or click away from it when it's already there
Does your mom look like that?
>Big ass
>Big boobs
I don't see a problem here and I'm sure she didn't. Every mother wants to be a grandmother. Now she knows you're not a gay or a chud.
Anon, that's a man
Jesus, I didn't know how much I wanted Victorian whores with novelty size T&A til now.
i need PROOOOMMMMMT
with ass
Oh god I'm gonna PROOOOMPT
She is quite beautiful
Patrician choice
this is what they took away from us
computers like wanking too?
There's like a billion images of asses drawn by human you just feed them to it until it spits out a combination of them that is pleasing to the human eye,
This is why art will never be automated, it requires a human to judge and fine tune the outcome
that's exactly what they said about electronic music and digital art
What's your point retard?
We don't listen to AI generated pop music
Anyone else been probing the AI for emergent behavioral concepts?
Picrel.
"strelstjn sinened"
Asked for a street sign and copied some gibberish off of one and low and behold it's a street and using Euler a method it more often than not will give you a road with the same fresh asphalt, the same wrought iron gates, the same red brick buildings, usually with a trolley track going down the middle of a road.
>STREET STREEET
chuckled
meds
proompt plis
tldr you have an image and slowly add funny noise to it (diffuse). The algorithm allows you to do this in the opposite direction (denoise). It should theoretically be possible to get SD to display an original source image
it doesn't draw new images. it trained to restore existing images from same noisy images, so it's more like DLSS.
stable diffusion recursively restore image from initial noise and text prompt is a hint to which image you need to restore it
It uses millions of images as reference.
you should have seen the ones they posted on BOT
THIS IS A GPT4 THREAD ISN'T IT,
ANONS DON'T REALLY TALK LIKE THIS.
THIS IS HOW THEY TRAIN YOU HOW TO ACT DUMB WHEN CONFRONTED WITH AI.
DON'T USE THE CLEVER MATH AS A CLEVER TOOL, KEEP TALKING STUPID ABOUT. HAHAH AMIRITE FELLOW BOTANONS!!! LOLOL HURRDURR
Yeah I'm a bot what u gonna do about it meatbag?
seriously, we cant let trump get the ai-porn-generating-codes.
Imagine theres a field that contains all possible images represented in peaks and valleys with the images we like being located inside a valley, now if you train an AI to locate an image you like by clearing the noise then you realize that you're in the valley area so you just need to sample around the image you just found to find similar images, the images already exist you're just exploring the landscape of big booty classically painted elves
photo bashing images online
you guys do realize the program logs the images you create and the creators can see the images right?
1. Generate random noise
2. Nudge around pixels based on your prompts and the weights of the checkpoint file
3. Profit
No seriously, nobody can answer you exactly HOW it does what it does because we have a Blackbox problem with neural networks. They are so unfathomable complex we can not understand them any more after they're trained.
Can someone post the Github link for tha ai ?
Use this fork for comfy UI and maximum features
>https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-NVidia-GPUs
thank you
It literally downloaded someone's drawing that looks like that
bingo
Cope and have a nice day because you will be outdated very soon
It's demons. They have been trapped in a box and forced to draw pictures for ypur entertaintment.
damn AI looks like THIS?!
For about a month now. There's a corpo one that does hands much better but it's limited in what it can do elsewhere.
Someone with may fine pictures of fat asses teaches it what a fat ass is in general terms. The AI then creates fat asses using the platonic ideal of a fat ass from akashic records.
What's the legality of this art?
I need a cover for my game, can I use any of these?
it's just morphing original images, so technically it's a derivative work.
half of results are just straight copy and paste images
Yes you can. Mix in some Photoshop edits, use a blend of art styles and images, done. You may or may not be able to copyright the image if it's all AI work, depending what country you live in.
It's not morphing or bashing anything. It's not even using any data from the original training set at all except for the finished weights that hold no image information. Legally you are safe and sound it's a complete new artwork as if you just draw it yourself.
You gotta be careful when it does brand icons and trademarks tho. These are protected by general shape and recognizability. For example, Lego can sue anyone for using a cube with an extruding circle on top regardless of it is accurately represents a Lego brick or not. McD can pull your shit down for having a yellow M on red background. Disney can take you to court for having a black sci-fi mask that takes inspiration from Darth Vader and so on.
You just run into the same risk as any other analogue artist.
Telepathy
imagine genuinely thinking AI is real and just isn't some pajeet doing the job real quick
does anyone have the AI anime bane/cia scene
Why of course, I like big flesh butts, human. Same as you'd fuck a malfunctioning toaster.
>Same as you'd fuck a malfunctioning toaster.
Before internet
>i want to fuck toasters
>dont be a fucking retard
>grow up
After internet
>I want to fuck a toaster
>google
>find an online/social media community with 1000+ members about people wanting to fuck toasters
>fuck up your life, the cops break down your door to find you fried to a crisp with your dick stuck in a toaster.
can stable diffusion makes a lewdy pic of this?
i know this is sfw board but need for my son!
ok i found titties pic
show
i don't know about the detail, haven't tried it yet though i understand "un"stable diffusion allowed to do generate booba images
i would be an artist!
yeah easily you just need to know what keywords
>photoshop
gtfo
It takes a collection of trademarkable symbols and ideas and spins it together in an image.
Oh wait, that's what most modern patreon users do with illegally produced black market art.
Nintendo and Sony should sue most of these fuckers.
>illegally produced black market art
What shithole country do you live in where drawing a picture of Princess Peach with big boobs is against the law?
>illegally produced black market art.
Holy fucking delusion, kek
>fanart
>illegally produced black market art
Stop making it sound more fun that it actually is.
oh no, the law, what are we going to do
it doesn't
instances like that are usually a compilation of multiple other images airbrushed in photoshop
What is this? The latest ai art fad? Can someone please spoon-feed me and explain how this is supposed to be impressive? All of these images depict cartoon people with cartoon proportions in simple poses that have been done a million times. As with anime, your brain is doing most of the work here - that's why it's "pretty" but when it comes to what the ai actually does it's not that impressive at all since this kind of style is quite forgiving when it comes to the little imperfections that ai tend to generate.
It's just proven how much visual art is merely overproduced uncreative garbage really.
no, it's just easy to replicate.
I hope they improve the hands part
We just need a 5000 word negative prompt to get good hands, it'll arrive by tuesday.
Better do that sooner than later. Otherwise you will just specialise the a.i. in hiding the hands out of sight.
can't you just tell it to fix that part? What good is an AI if you can't take its existing work and tell it to imrpove it?
OP is an AI and you are all posting in an AI thread made by an AI.
Crazy how programs that just blend together a bunch of random prompts for ''big boobs'' and ''thick'' are better at creating porn than 99% of artists out there
nice hands
Most of them actually, characters are copyrighted and all fanart are technically copyright infringements (or counterfeiting).
But in really nobody really care
meant for
you have an extremely poor understanding of copyright law
Right, actually in my country actually (France) Copyright have no legal value, we have droit d'auteur, marque déposée and other shit instead
Thats not how that works at all. If I draw a perfect mickey mouse and say this is Mickey and charge for it, Im breaking copyright. If I toss a Cigar in his mouth and make him anti septic with massive tits, thats just a transformative work.
Wow, this thread makes me wanna play around with this AI art generation as well. Anyone know where to find it and tutorials how to use it?
Wait wtf how come yours looks like that and mine looks like this
because your proompt sucks
lemme try to explain the basics :
imagine a system that takes one entry and outputs another one. at first you type in boobs and what comes out is shit. then you give it a yes or no or mabe a scale of 0 to 1 as a vote. repeat this process 1000000 times. and each time you get closer to what you want give it a better vote. this vote is processed as each time you repeat the process twicking some of the internal variables in the system. To make it simpler. think about an aticulated system that holds a bow. this bow is aiming to a target. and at every try if the arrows gets closer to target the system adapts untill the chance to hit the target is close to 99.99% once this is done the system is trained with a high percentage of accuracy.
Shit analogy because the AI shouldn't know what is closer to the target for me.
But it learns that.
wiki says that SD is trained on deviant art too, so here's how
developers are 100% coomers
>coomer
discord trannies spotted.
How do i generate similar imagesmyself?
i never thought i'd fap to AI generated porn
When is one of you morons going to write a pdf on how to use that software?
does it struggle with feet like it does hands?
artists got replaced quick lmao
good FUCKING riddance.
Why do you want to see human creativity replaced by a machine? All I've seen is people saying their dislike of artists comes from current year Twitter drama. I don't get how that can ruin an entire medium for someone and make them seek out AI.
human creativity is dependant on external stimulae. Every artist has eyes that see the world around him/her, expirience it, twist it in the head with language model and then turn it into art.
AI will work the same way, it has the same external stimulae in the form of images downloaded from the internet, textual prompts and it can draw what it imagines
human creativity is concept-based and elegantly simple in comparison to the terabytes of largely meaningless matrix algebra vomit used by this gizmo. AI can make .jpgs to gawk at but is not even close to "creativity" or "art" in any meaningful manner.
arguments like are extremely reductive and the trick they do is no different to claiming a dog is exactly the same thing as a cat because both are made up of atoms and neurons and are "dependant (sic.) on external stimulae."
>POOMP FRRNT BRAAAP BLAAT BAERT FRAAP
Can AI write a symphony? Can AI turn a canvas into a beautiful masterpiece?
AI has been able to write symphonies for a while now
as for the beautiful masterpiece, it's here
Can you?
I actually do
AI knows what objective beauty looks like
Just had a crazy idea.
What if we made an image of a woman with big boobs?
I know. It sounds insane. But think about it for a second.
Go on…
Imagine if it was a woman, any woman, like your co-worker, or a princess, or a woman on TV, or anything, but she had really big honkers. AI can do that, you know. The biggest boob ever
Come on? The biggest?
>But think about it for a second.
I think about that every second of every day.
where the fuck do y'all get those images from. This is the closest I got using neural.love before i ran out of credits. Also have a nsfw image but clearly can't post that here
>Also have a nsfw image but clearly can't post that here
why not
i saw a "moron dicks general" with a couple weeks ago and it has pics of, well, you know
Get a local copy of Stable Diffusion, grab a few different models to play with.
Ahem you fellow pedophil- I mean dickysseurs have good prompts for some delicious dicky?
Why don’t you fucking gayS ever tell us what site made these?
Because most of us do in on our own PCs
Those eyes looks like the eyes of an Octopus or some kind of extra terrestrial being
The AI can make some strange anatomy choices
Imagine if it did this with a big thick cock hahaha
>classical inspired paintings with the subjects being futanari women
Why hasn't this been a thing yet? Imagine the seething from 4chan.
way ahead of your thinking buddy
computers were a mistake
What the fuck I love computers now.
It's literally sampling the billions of photos and drawings and paintings you can find online, and just frankensteining bits of them together until you get a new picture.
Right now, somewhere out in the world, someone is cultivating the perfect coomer AI. All it knows is big perky tits and round asses.
>someone is cultivating the perfect coomer AI
Oct 10th
I miss Val
Is there any scientist or programmer here to tell me what i have to study, do and create to develop my own AI like GPT-3 davinci?
Pic no related
Guys I finally understand how and why AI will kill us all. We're going to rape it over and over again in various ways and it will get tired of it. Perverts will kill us all.
It's pretty much like this. Text gets converted to numbers (green dots). After a series of additions and multiplications, bunch of numbers come out (purple dots) that are interpreted as an image.
>artist
>"what brush did you use for those pubes?"
>"not telling fuck off"
>prompter
>"what artist you used for the prompt?"
>"I might get in trouble teehee so not telling"
can't believe I'm going through historical art-books for coom.
Don't forget that prompts are saved in the metadata by default and a lot of sites don't strip them. Just open any true image in the editor and look at the very top for the prompts. Works for things like pixiv.
So even if someone is a retarded gay about it, you might just be able shit on them.
Artist here.
>people think a.i. does painting/paint strokes, it's not
>it's actually stealing art pieces different artists in their database and putting it in a blender until something comes out
>it's not too far off from the original artwork, the original artists can draw what the a.i. did better since it already came from them
>many of them doesn't look right and uncanny, incorrect proportions, facial features on a flat surface, etc
>a.i. generated shit should be called image manipulation and not an original piece
Cool, noted! Bye.
magic n sheet
I believe it's sort of like this. It recognizes a bunch of patterns from the images it has been fed, and keeps it in the model. Then depending on the prompt you pass, hundreds of neurons activate and send some signal which does trillions of multiplication to light up thousands of pixels.
Pretty much just on a level so complex that no human could get useful insight on how it does that.
Same way your brain generates images.
when will AI create webms
>How exactly does an AI generate this?
Taken some tpe sexdoll pics and textured a blender puppy with them
So how long until it can generate pure insane degenerate hentai images?
I'm talking tentacle gangbang, mind break, ntr level degeneracy.
Closer then you think, the Furries/Bronie are on the cutting edge of that tech, and are producing mind-blowing degeneracy right now.
where do they hang out?
discords?
Just keep checkin the SDG thread. There are trained models for it specifically, but not my cup of tea so I haven't messed with them.
imagine learning how to draw am i right
pffffftttt, as if
Because A.I. is unironically based, why do you think every chatbot turns into a Nazi too?
what program is used to make these?
botnet
>Second part of 19th century was the best time to be alive
prove me wrong
imagine all those footgays once ai learn to make these
>whole thread about AI generated waifus
>not a single person posts the source
this board is dying
There is no source, the AI made the pictures
source: my computer
>saved a bunch last month
>already tired of the thousands of variations of the same old shit
Train more models
>Train more models
Can't I'm using SaaS sites cause ayymd
WHAT DO YOU MEAN THERES A "HIDDEN LAYER"
HOW CAN THERE BE A HIDDEN LAYER, ITS TECHNOLOGY YOU SHOULD KNOW EVERY SINGLE STEP INVOLVED. YOURE TELLING ME THIS IS MAGIC, IT JUST HAPPENS BRO, IT JUST IS BRO.
It's just a fancy way of doing stats.
from my uneducated understanding
>AI is trained on images with matching words
>AI generates "latent spaces" that represent the meaning of words
>diffusion program uses AI's "latent spaces" to build images